AI strategy

Most of what decides an AI project is not technical. The readiness checklist turns that into questions and a score: fewer than five you cannot answer, start with a prototype on real data; five to ten, run a short assessment; more than ten, the data is the first project. The MCP decision guide applies the same thinking to a question product teams now face, whether to let AI agents into the product, with five signals to build now and three to wait.

The checklist comes from our AI strategy and readiness assessment, a one-to-two-week engagement run by the engineers who would build the system. It produces an opportunity map, the data and guardrail requirements, two or three architecture options with their monthly running costs, a prioritised build plan and a prototype scope. When a provider changes its model, an evaluation set with known answers, run before and after, is the only way to know whether results got better or worse. For how this plays out in practice, see MayAI, a consumer agent taken from a chat message to a working prototype.